js string should use `Uint8Array` typed arrays

[timotheecour]

links

• https://stackoverflow.com/questions/6965107/converting-between-strings-and-arraybuffers
• https://www.npmjs.com/package/@fictivekin/growable-uint8-array
• https://github.com/nim-lang/Nim/pull/16461
• see also:

  proc arrayTypeForElemType(typ: PType): string =
  # XXX This should also support tyEnum and tyBool
  case typ.kind
  ...
  of tyInt8: "Int8Array"
  of tyUInt8: "Uint8Array"
  of tyFloat64, tyFloat: "Float64Array"
  else: ""

note

insome benchmakr i did though, TextEncoder.encode seems slow.

• 4383 - TextEncoder.encode is slow - v8

Ref https://github.com/timotheecour/Nim/issues/156

[timotheecour]

• [ ] why does cstrToNimstr differs from this implementation?

in particular this part:

// Surrogate Pair
      c = 0x10000 + ((c & 0x03FF) << 10) + (str.charCodeAt(++i) & 0x03FF);

seems different?

https://github.com/google/closure-library/blob/master/closure/goog/crypt/crypt.js#L110

/**
 * Converts a JS string to a UTF-8 "byte" array.
 * @param {string} str 16-bit unicode string.
 * @return {!Array<number>} UTF-8 byte array.
 */
goog.crypt.stringToUtf8ByteArray = function(str) {
  'use strict';
  // TODO(user): Use native implementations if/when available
  var out = [], p = 0;
  for (var i = 0; i < str.length; i++) {
    var c = str.charCodeAt(i);
    if (c < 128) {
      out[p++] = c;
    } else if (c < 2048) {
      out[p++] = (c >> 6) | 192;
      out[p++] = (c & 63) | 128;
    } else if (
        ((c & 0xFC00) == 0xD800) && (i + 1) < str.length &&
        ((str.charCodeAt(i + 1) & 0xFC00) == 0xDC00)) {
      // Surrogate Pair
      c = 0x10000 + ((c & 0x03FF) << 10) + (str.charCodeAt(++i) & 0x03FF);
      out[p++] = (c >> 18) | 240;
      out[p++] = ((c >> 12) & 63) | 128;
      out[p++] = ((c >> 6) & 63) | 128;
      out[p++] = (c & 63) | 128;
    } else {
      out[p++] = (c >> 12) | 224;
      out[p++] = ((c >> 6) & 63) | 128;
      out[p++] = (c & 63) | 128;
    }
  }
  return out;
};

edit: see this other implementation: https://stackoverflow.com/a/64277403/1426932 (From emscripten)

[ringabout]

I agree. I think cstring should use uint8Array.

[AmjadHD]

Will this simplify makeNimstrLit, cstrToNimstr and makeJSStr ?